DO%sat & DO mass:

Figure 1. Volume weighted dissolved oxygen percent saturation and total dissolved oxygen mass.

Defining periods of comparison:

Figure 2. We are interested in comparing the effects of treatment against the Sparkling reference and untreated Crystal condtions. To do this we subdivide each year of lake observation into three periods “Spring”, “Summer”, “Autumn”, where in untreated Crystal years (and all years of Sparkling Lake) “Spring” is ice-off to the onset of thermal stratification, “Summer” is the start of stratification to the end of stratification, and “Autumn” is the end of stratification to mid-November (ice-on is typically late-December in Crystal (treatment) and Sparkling (reference) Lakes). The beginning of stratification is the day when Schmidt Stability first exceeded 162 J/m^2 for more than 5 days before the annual stability maximum. The end of stratification is the day when Schmidt Stability has first decreased below 20 J/m^2 for more than 30 days after the annual stability maximum. We choose a threshold of 162 J/m^2 because this is the average Schmidt Stability corresponding to the onset of thermal stratification in Crystal Lake in treatment years as defined as the date when the sum of squared differences between treatment and reference lake water temperatures (1 - 18 m; 1 m resolution), was maximized in 2011 and 2014. In otherwords, the date when the effects of GELI treatment on water temperatures were beyond that “experienced during untreated conditions, corresponded with a Schmidt Stability of 175 J/m^2 and 150 J/m^2 in Crystal Lake for 2012 adn 2013 respectively. We then used this Schmidt Stability value to define the onset of thermal stratification in both treatment and reference lakes and in each year of the experiment (2011-2014). A lower value of 20 J/m^2 was choosen to mark the end of stratification.

NOTE: Not all figures posted below use this definition for grouping into periods (I haven’t had time to update them all yet). Figure captions will tell you which grouping method was used.

NEP (6 hr):

Figure 3. Six-hour depth integrated NEP (grey points) and 3 day moving median (red line). Missing data was filled with the monthly average before smoothing. NEP = delta.O2 - flux.O2, where delta.O2 is the change in the areal whole lake dissolved oxygen between consecutive 6-hour profiles, and flux.O2 is the O2 flux between the atmosphere and lake calculated every minute and integrated over the 6 hour period between profiles. The k.gas model used for calculation is Cole and Caraco 1998.

Figure 4. Six-hour NEP.di (orange points) and NEP.uml (grey points).

Figure 5. Annual averages of 6-hour depth integrated NEP grouped to 6-hour interval. Error bars denote standard error not corrected for autocorrelation.

Figure 6. Six-hour depth integrated NEP grouped to 6-hour period for each year.

Figure 7. Annual averages of 6-hour depth integrated NEP grouped to each 6-hour interval and to season. Error bars denote standard error not corrected for autocorrelation. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November.

Figure 8. Six-hour depth integrated NEP grouped to 6-hour interval and to season. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November.

Figure 9. Upper mixed layer NEP vs. depth integrated NEP for the full open water period of each year. NEP is of 6-hour resolution. Dashed line is 1:1. Solid red line is the type II major axis regression line bounded by confidence intervals. Regression slope, R^2, and correlation coefficient are presented.

Figure 10. NEP.uml vs. NEP.di grouped by season and year. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. NEP is of 6-hour resolution. Dashed line is 1:1. Solid red line is the type II major axis regression line bounded by confidence intervals. Regression slope, R^2, and correlation coefficient are presented.

Table 1. Slopes of type II major axis regression for NEP.uml vs. NEP.di for each daily six-hour sub-interval of each season and year (Figure 11, below). BLUE = six hour interval in which the slope is closest to 1:1 within each season of each year. RED = six-hour interval in which the slope is furthest from 1:1 within each season of each year.

Figure 11. Six-hour NEP.uml vs. NEP.di grouped by 6-hour interval and season. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. Data point colors accord with previous figures. Dashed line is 1:1. Solid red line is the type II major axis regression line bounded by confidence intervals. Regression slope, R^2, and correlation coefficient are presented.

NEP (daily):

Figure 12. Daily NEP.di (orange points) and NEP.uml (grey points).

Figure 13. Daily NEP.DI (grey points) and 18 day centered moving mean (red line). Time series gaps have been filled with daily NEP.DI estimates obtained from interpolated then smoothed depth integrated oxygen measurements.

Figure 14. Moving average and moving standard error (18 day centered) of daily NEP.DI. Gaps have been filled with daily NEP.DI estimates obtained from interpolated and smoothed depth integrated time series.

Figure 15. Daily NEP.UML vs. NEP.DI grouped by season. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. Dashed line is 1:1. Solid red line is the type II major axis regression line bounded by confidence intervals. Regression slope, R^2, and correlation coefficient are presented.

NEP (weekly):

Figure 16a. “Weekly”" (six day) depth integrated NEP estimated each day from a rolling window that calculates weekly NEP as: NEP = delta.DO - gas.flux. Where, delta.DO is the difference between DO concentrations +3days and -3days, and gas.flux is integrated minute resolution gas flux over the same period. Fig. 16b provides some context to these estimates. NOTE x-axes scales of Figs 16a and 16b are different.

Figure 16b. Whole lake volume weighted dissolved oxygen percent saturation.

Figure 16c. Means and standard error of “weekly” depth integrated NEP estimated each day from a rolling window (see Fig. 16a caption) grouped into seasons (defined above in section “Defining periods of comparison”). Standard error has not been corrected for autocorrelation. Blue = 2011, red = 2012, orange = 2013, green = 2014. Number of samples for each year and within each season are listed at the bottom of the figure. One-way ANOVA, on each season, revealed statistically significant differences. Tukey-Kramer post-hoc test revealed non-significant differences between years (within a season) noted by letters.

Metabolism (daily):

Figure 17. Daily GPP (green), R (brown), NEP (grey).

Figure 18. Average daily GPP (green), R (brown), NEP (grey), grouped to stratification phenology. Values beneath bars correspond to means. Confidence intervals are standard error, not corrected for auto correlation. Upper mixed layer data is presented in the top pannel and depth integrated is presented in the bottom pannel. Periods of stratification are defined in the section “Defining periods of comparison”.

Figure 19. Average annual GPP (green), R (brown), NEP (grey), grouped to stratification phenology. Values beneath bars correspond to means. Confidence intervals are standard error, not corrected for auto correlation. Upper mixed layer data is presented in the top pannel and depth integrated is presented in the bottom pannel.

GPP (daily):

Figure 20. Daily GPP.uml vs. GPP.di for the full open water period. GPP.uml and GPP.di are estimated from 06:00-18:00 (NEP day), and 00:00-06:00 (NEP night). Negative values of GPP have been omitted. Solid line is the type II major axis regression. Dashed line is one to one. Regression equation and correlation coefficient are presented.

Figure 21. Daily GPP.uml vs. GPP.di grouped by season. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. GPP is of daily resolution. Negative values of GPP have been omitted. Solid line is the type II major axis regression. Dashed line is one to one. Regression equation and correlation coefficient are presented.

R (daily):

Figure 22. Daily R.uml vs. R.di for the full open water period. R.uml and R.di are estimated from during 00:00-06:00 (NEP night). Dashed line is one to one. Days of bad estimates have been omitted, where “bad” estimates are considered R > 0. Solid line is the type II major axis regression. Regression equation and correlation coefficient are presented.

Figure 23. R.uml vs. R.di grouped by season. R is of daily resolution. Days of bad estimates have been omitted, where “bad” estimates are considered R < 0. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. Solid line is the type II major axis regression. Dashed line is one to one. Regression equation and correlation coefficient are presented.

Gas flux:

Figure 24. Crystal Lake gas flux minute resolution.

PAR:

Figure 25. Daily average PAR.

Figure 26. Daily average water column PAR grouped by season. Spring = ice-off to summer solstice; Summer = summer solstice to fall equinox; Fall = fall equinox to mid-November. Standard error is not corrected for autocorrelation. Colors correspond to year as in previous plots (blue = 2011, red = 2012, orange = 2013, green = 2014).

z.metalimnion, z.photic:

Figure 27. Dissolved oxygen percent saturation, metalimnion top (grey), metalimnion bottom (black), and depth of the photic zone (yellow; 1% surface PAR).

Figure 28. Chlorophyll a fluoresence (ug/L), metalimnion top (grey), metalimnion bottom (black), and depth of the photic zone (yellow; 1% surface PAR).

BACI:

Figure 29. Time series difference (Crystal - Sparkling) of daily UML lake metabolism for each year. Respiration was converted to positive values before subtraction. Note that a large percent of daily metabolism estimates were excluded from consideration. Metabolism estimates were excluded when daily R > 0 or GPP < 0. Daily NEP was removed on these days of “bad” GPP and R. Percent daily metabolism removed because of “bad” GPP or R:

Lake.Year % removed
CR.2011 22%
CR.2012 20%
CR.2013 21%
CR.2014 25%
SP.2011 50%
SP.2012 53%
SP.2013 54%
SP.2014 34%

Figure 30. Means and standard error of daily UML metabolism difference (Crystal - Sparkling) grouped into seasons (defined above in section “Defining periods of comparison”). Standard error has not been corrected for autocorrelation. Blue = 2011, red = 2012, orange = 2013, green = 2014. Number of samples for each year and within each season are listed at the bottom of the figure. One-way ANOVA, on each season, revealed statistically significant differences. Tukey-Kramer post-hoc test revealed non-significant differences between years (within a season) noted by letters.